Recording Interactive Lectures

[Bennyp]

Hello Friends,
Thank you for considering my question.

I record lectures on a regular basis. My dream is to have a system which produces clean audio files with two tracks: one of the lecturer at optimal volume levels, and another of the audience at legible levels.

The lecturer sits at a desk in front of a small library, about 6-7 meters square. About 11 students sit at desks opposite him, in 3 rows which fill the space.

• Samson CO2 Mics on a stereo Bar
• Tascam DR-100

I'm using a stereo-matched pair of "Samson CO2" mics (I know), screwed into a stereo bar and held up using a regular mic stand. The mics have a cardioid pattern. The Left channel mic is pointed up at a 45 degree angle towards the speaker. the Right channel mic is pointed 180 opposite, 45 degree angle down.

The lecturer comes out loud and clear, and sounding quite nice, compared to the average voice recorder. The audience, however comes out too quietly. I have experimented with using a higher input level for the right channel (audience) mic, but I didn't like the results, as I was getting too much reflection of the speaker, who often speaks louder than the audience.

So the trade off I have to make is between hearing the audience's questions and wrecking the lecturer's signal with reflections.

Any tips on mic type/placement or clever processing techniques to deaden the reflections and accentuate the questions?

Thanks again for reading,
BP

 

[watkinsp]

Hi BP,

I have exactly the same question. I haven't found much info on this topic on the web. I'm also using omnis for audience questions and putting the audience questions on one track with the speaker on another. I'm planning on using a compressor with a sidechain input to lower the volume of the audience mics whenever the speaker talks. Unfortunately, I still have the problem of ambient noise in the audience mics (HVAC noise mostly).

If anyone else has any tips, I'd appreciate it.

 

[watkinsp]

So I used the compressor to duck the audience when the speaker was talking. It worked, but it sounded kinda like C-SPAN where you can hear the speaker talk and then the roar of the conversations in the house intermittently. I'll still have to do post processing later.

I tried a noise cancelling scheme using two matched mics and inverting the output of one. This worked for canceling a low frequency sine wave, worked slightly for canceling a square wave, and just didn't work at all for canceling white noise (similar to HVAC noise). So it looks like some sort of noise cancelling setup isn't going to eliminate HVAC noise.

The other thing I'm going to try is distributing four wireless Karaoke mics around the audience. When someone wants to ask a question, he or she will have to talk into one of the mics.

 

[watkinsp]

I've come up with an idea to gate the HVAC noise out of my audience mics. The gist of it is: Only let the audience audio through when the presenter isn't talking and the audience is.

To determine when the audience is talking, I'll use a narrow bandpass filter. This should mostly eliminate HVAC noise and let through a very low-fi signal of just the audience and presenter speaking. This signal will drive the sidechain input of a compressor on the audience mics. The output of the compressor is the audience audio (with HVAC noise) except that the audio is muted when the audience (or the presenter) is talking. A mixer combines this signal and the presenter audio. The output of the mixer is only silent when the audience alone is talking. This signal goes to the sidechain input on a second compressor (which has the original audience mics as its input). The output of the second compressor is the audience mics muted at all times except when the audience is talking.

I've sketched up a diagram below:


The next step is to run some tests in MatLab with some audio clips I've already recorded. Let me know if you think it won't work (and why).